shep@datacube.UUCP (07/06/85)
A new architecture, MaxVideo's MAXbus, and a hardware module, VFIR (Video Finite Impulse Response filter) allow 1-d and 2-d convolutions to be performed at real-time speeds. A linear growth in kernel size requires linearly more hardware or more time. 10*10 non-separated convolutions against a 512*512*8 image takes 1/3 second with a single board. An exact product is result. With this fast hardware facility, we are looking for applications code to generate/manipulate the kernel(s) for image processing and machine vision applications. Pointers to texts, papers, or existing product, or people would be appreciated. Filter design: 1. Application of remex exchange equiripple calculations to a 2-d kernel. Possible specifing frequency windows along polar co-ords? 2. Application of window design techniques (Hamming, Hanning, Lanczos, etc). Ideally 'C' code for the 2-d method from Cappellini's text. Filter analysis: 1. Code to plot frequency and amplitude in 3-d of a 2-d kernel. If you really like to code this stuff, you may even want to send a resume too...(!) Shep Siegel ihnp4!datacube!shep Datacube Inc. ima!inmet!mirror!datacube!shep 617-535-6644 decvax!cca!mirror!datacube!shep 4 Dearborn Rd. decvax!genrad!wjh12!mirror!datacube!shep Peabody, Ma. 01960 {mit-eddie,cyb0vax}!mirror!datacube!shep